I'm currently implementing a setup that includes 2 CRS boxes with one CGSE+ each. Each one of these CRSs has its own uplinks to the Internet. The method I'm planning on using to inject traffic into each CGSE is MPLS VPN, with CRS-A injecting a default route into the VPN with better metrics than the one injected by CRS-B. This mechanism would force traffic to traverse NAT-A in CRS-A and in case of failure on CRS-A, routing would force traffic into CRS-B so it get NAT'd there.
My question has to do with the ip pools I'm going to configure for inside-vrf to be NAT'd on each CRS. I have 2 options:
1) using different pools: traffic would normally pass through CRS-A and get NAT'd into pool-A. in case of failure, routing would force traffic through CRS-B and would get NAT'd into pool-B
2) using the same IP pool on both CRSs:traffic would normally pass through CRS-A and get NAT'd into pool-X. in case of failure, routing would force traffic through CRS-B and would get NAT'd into pool-X.
Are both setups valid? I mean would both work and provide high availability? Is one recommended over the other?
Reason I ask is customer really needs to have as many IPv4 addresses as possible available. option 1 would require having pool B (under normal conditions) doing nothing most of the time. So, I'd really like to implement option 2 if possible (recommended).